A design pattern that connects a vision encoder to a language model, enabling the language model to understand and describe images.